Akaike information criterion should not be a “test” of geographical prediction accuracy in ecological niche modelling

Velasco, Julián A.; González‐Salazar, Constantino

doi:10.1016/j.ecoinf.2019.02.005

Cited by 81 publications

(54 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…More specifically, to obtain the most accurate representation of Leptospira circulation in the landscape, it would be necessary to assess the presence of Leptospira serovars in wildlife and the environment to provide and integrative estimation of the geographic and environmental risk (Albert, Goarant, & Mathieu, ). The second limitation that we found was related to the method used for the estimation of the ENM performance, since a recent paper (Velasco & González‐Salazar, ) highlighted the need to reduce the use of AIC for geographic predictions in ecological niche modelling studies due to the lack of accuracy of this approach on evaluating model performance, which is currently one of the most used methods to estimate ENM complexity (Cobos, Peterson, Barve, & Osorio‐Olvera, ; Escobar, Qiao, Cabello, & Peterson, ; Freeman, Sunnarborg, & Peterson, ; Raghavan, Peterson, Cobos, Ganta, & Foley, ).…”

Section: Discussionmentioning

confidence: 99%

Spatial distribution and spread potential of sixteenLeptospiraserovars in a subtropical region of Brazil

Jara

Escobar

Rodriges³

et al. 2019

Transbound Emerg Dis

View full text Add to dashboard Cite

Leptospirosis is a bacterial disease that represents a major problem in animal and public health due to its high prevalence and widespread distribution. This zoonotic disease is most prevalent in tropical environments where conditions favour pathogen survival. The ecological preferences of Leptospira serovars are poorly understood, limiting our knowledge of where and when outbreaks can occur, which may result in misinformed prevention and control plans. While the disease can occur consistently in time and space in tropical regions, research on the ecology of leptospirosis remains limited in subtropical regions. This research gap regarding Leptospira ecology brings public and veterinary health problems, impacting local economies. To fill this gap of knowledge, we suggest to assess geographic and ecological features among Leptospira serovars in a subtropical area of Brazil where leptospirosis is endemic to (a) highlight environmental conditions that facilitate or limit Leptospira spread and survival and (b) reconstruct its geographic distribution. An ecological niche modelling framework was used to characterize and compare Leptospira serovars in both geographic and environmental space. Our results show that despite the geographic overlap exhibited by the different serovars assessed, we found ecological divergence among their occupied ecological niches. Ecological divergences were expressed as ranges of potential distributions and environmental conditions found suitably by serovar, Sejroe being the most asymmetric (<0.15). Most important predictors for the potential distribution of most serovars were soil pH (31.7%) and landscape temperature (24.2%). Identification of environmental preferences will allow epidemiologists to better infer the presence of a serovar based on the environmental characteristics of regions rather than inferences based solely on historical epidemiological records. Including geographic and ecological ranges of serovars also may help to forecast transmission potential of Leptospira in public health and the food animal practice.

show abstract

Section: Discussionmentioning

confidence: 99%

Spatial distribution and spread potential of sixteenLeptospiraserovars in a subtropical region of Brazil

Jara

Escobar

Rodriges³

et al. 2019

Transbound Emerg Dis

View full text Add to dashboard Cite

show abstract

“…We used the package ENMeval in R v. 3.6.1 [50] to optimize the MaxEnt model, set the regulatory multiplier (RM) to 0.5-8, and each interval was 0.5, for a total of 16 regulatory multipliers [48,51]. We used 9 feature combinations (FCs): L, LQ, H, LQH, LQHP, LQHPT, QHP, QHPT, and HPT(The MaxEnt model provides 5 features, which are linear features (L), quadratic features (Q), segmented features (H), product features (P), and Threshold features (T)) [52,53]. The ENMeval data package was used to test the above 144 parameter combinations, and we finally used the Akaike Information Criterion (AIC) model of the Akaike information criterion, and used 10% training omission rate (OR 10 ) and the difference between the AUC values (AUC DIFF ) to check the fit and complexity of the model (Table S2) [50].…”

Section: Model Establishment Optimization and Evaluationmentioning

confidence: 99%

Distribution Pattern of Endangered Plant Semiliquidambar cathayensis (Hamamelidaceae) in Response to Climate Change after the Last Interglacial Period

Zhao

Zhang

et al. 2020

Forests

View full text Add to dashboard Cite

Semiliquidambar cathayensis is a special and endangered plant in China, used for traditional Chinese medicine and in landscape applications. Predicting the impact of climate change on the distribution of S. cathayensis is crucial for its protection and the sustainable use of resources. We used the maximum entropy (MaxEnt) model optimized by the ENMeval data packet to analyze the potential geographic distribution changes of S. cathayensis in 12 provinces of Southern China for the different periods since the last interglacial period (LIG, 120–140 ka). Considering the potential geographic distribution changes in the province, and based on the two climate scenarios of Representative Concentration Pathways (RCP) 2.6 and RCP 8.5, the distribution range of S. cathayensis was analyzed and we predicted the range for the 2050s (average for 2041–2060) and 2070s (average for 2061–2080). The area under AUC (Area under the receiver operating characteristic (ROC) curve) is 0.9388 under these parameters, which indicates that the model is very accurate. We speculate that the glacial period refugia were the Nanling and Wuyi Mountains for S. cathayensis, and central and Western Fujian and Taiwan are likely to be the future climate refugia. In the mid-Holocene (MH, 6 ka), the growth habitat was 32.41% larger than the modern habitat; in the 2050s and 2070s (except RCP2.6–2070s), the growth habitat will shrink to varying degrees, so efforts to support its in situ and ex situ conservation are urgently needed. The jackknife test showed that the main factors affecting the geographical distribution of S. cathayensis were annual precipitation, precipitation of the wettest month, and precipitation of the driest month. The annual precipitation may be the key factor restricting the northward distribution of S. cathayensis. In general, the centroid of the distribution of S. cathayensis will move northward. The centroid of the adaptive habitats will move northward with the highest degree of climate abnormality. We think that Hainan Island is the most likely origin of S. cathayensis. These findings provide a theoretical basis for the establishment of genetic resources protection measures, the construction of core germplasm resources, and the study of the formation and evolution of Hamamelidaceae.

show abstract

“…This result suggests that the quality of the model can have an effect on the accuracy of the resulting land-cover classifications, but this effect could largely differ depending on the land-cover under analysis, the threshold used for building the binary maps and the metric used for evaluating the quality of the model [ 16 ]. Therefore, while for some cases the quality of the model has a positive effect on the classification accuracy (e.g., [ 27 ]), in other situations, there may be no relation between them (e.g., [ 28 ]).…”

Section: Discussionmentioning

confidence: 99%

Land-Cover Classification Using MaxEnt: Can We Trust in Model Quality Metrics for Estimating Classification Accuracy?

Morales

Fernández

2020

Entropy

View full text Add to dashboard Cite

MaxEnt is a popular maximum entropy-based algorithm originally developed for modelling species distribution, but increasingly used for land-cover classification. In this article, we used MaxEnt as a single-class land-cover classification and explored if recommended procedures for generating high-quality species distribution models also apply for generating high-accuracy land-cover classification. We used remote sensing imagery and randomly selected ground-true points for four types of land covers (built, grass, deciduous, evergreen) to generate 1980 classification maps using MaxEnt. We calculated different accuracy discrimination and quality model metrics to determine if these metrics were suitable proxies for estimating the accuracy of land-cover classification outcomes. Correlation analysis between model quality metrics showed consistent patterns for the relationships between metrics, but not for all land-covers. Relationship between model quality metrics and land-cover classification accuracy were land-cover-dependent. While for built cover there was no consistent patterns of correlations for any quality metrics; for grass, evergreen and deciduous, there was a consistent association between quality metrics and classification accuracy. We recommend evaluating the accuracy of land-cover classification results by using proper discrimination accuracy coefficients (e.g., Kappa, Overall Accuracy), and not placing all the confidence in model’s quality metrics as a reliable indicator of land-cover classification results.

show abstract

Akaike information criterion should not be a “test” of geographical prediction accuracy in ecological niche modelling

Cited by 81 publications

References 42 publications

Spatial distribution and spread potential of sixteenLeptospiraserovars in a subtropical region of Brazil

Spatial distribution and spread potential of sixteenLeptospiraserovars in a subtropical region of Brazil

Distribution Pattern of Endangered Plant Semiliquidambar cathayensis (Hamamelidaceae) in Response to Climate Change after the Last Interglacial Period

Land-Cover Classification Using MaxEnt: Can We Trust in Model Quality Metrics for Estimating Classification Accuracy?

Contact Info

Product

Resources

About